N.L., R.-X.C. and Y.S. contributed equally to this work
Early Detection and Diagnosis
A four-miRNA signature identified from genome-wide serum miRNA profiling predicts survival in patients with nasopharyngeal carcinoma
Article first published online: 30 SEP 2013
© 2013 UICC
International Journal of Cancer
Volume 134, Issue 6, pages 1359–1368, 15 March 2014
How to Cite
Liu, N., Cui, R.-X., Sun, Y., Guo, R., Mao, Y.-P., Tang, L.-L., Jiang, W., Liu, X., Cheng, Y.-K., He, Q.-M., Cho, W. C.S., Liu, L.-Z., Li, L. and Ma, J. (2014), A four-miRNA signature identified from genome-wide serum miRNA profiling predicts survival in patients with nasopharyngeal carcinoma. Int. J. Cancer, 134: 1359–1368. doi: 10.1002/ijc.28468
Conflicts of interest: Nothing to report
- Issue published online: 6 JAN 2014
- Article first published online: 30 SEP 2013
- Accepted manuscript online: 2 SEP 2013 08:58AM EST
- Manuscript Accepted: 27 AUG 2013
- Manuscript Received: 22 MAY 2013
- Innovation Team Development Plan of the Ministry of Education. Grant Number: IRT1297
- National Natural Science Foundation of China. Grant Number: 81201746
- Key Scientific and Technological Innovation Program for Universities of Guangdong Province. Grant Number: cxzd1005
- Science and Technology Project of Guangzhou city. Grant Number: 12C22061586
- Key Laboratory Construction Project of Guangzhou City, China. Grant Number: 121800085
- Guangdong Province Universities and Colleges Pearl River Scholar Funded Scheme (2010)
- nasopharyngeal carcinoma;
Recent findings have reported that human serum microRNAs (miRNAs) can be used as prognostic biomarkers in various cancers. We aimed to explore the prognostic value of serum miRNAs in nasopharyngeal carcinoma (NPC) patients. The level of serum miRNA was retrospectively analyzed in 512 NPC patients recruited between January 2001 and December 2006. In the discovery stage, a microarray followed by reverse transcription-quantitative polymerase chain reaction was used to identify differentially altered miRNAs in eight patients with shorter survival and eight patients with longer survival who were well matched by age, sex and clinical stage. The identified serum miRNAs were then validated in all 512 samples, which were randomly divided into a training set and a validation set. Four serum miRNAs (miR-22, miR-572, miR-638 and miR-1234) were found to be differentially altered and were used to construct a miRNA signature. Risk scores were calculated to classify the patients into high- or low-risk groups. Patients with high-risk scores had poorer overall survival [hazard ratio (HR), 2.54; 95% confidence interval (CI), 1.57–4.12; p < 0.001] and distant metastasis-free survival (HR, 3.28; 95% CI, 1.82–5.94; p < 0.001) than those with low-risk scores in the training set; these results were confirmed in the validation and combined sets. The miRNA signature and TNM stage were independent prognostic factors. The combination of the miRNA signature and TNM stage had a better prognostic value than the TNM stage or miRNA signature alone. The four-serum miRNA signature may add prognostic value to the TNM staging system and provide information for personalized therapy in NPC.
According to the International Agency for Research on Cancer, there were an estimated 84,400 incident cases of nasopharyngeal carcinoma (NPC) and 51,600 deaths in 2008. Globally, NPC has an extremely unbalanced endemic distribution, and the age-standardized incidence per 100,000 males ranges from 20−50 in Southern China to 0.5 in predominantly Caucasian populations.
Prognostic assessment is critical for making better therapeutic choices for NPC patients, and the tumor-node-metastasis (TNM) staging system is the key prognostic determinant. However, large variability in disease outcomes has been observed in a subset of NPC patients with the same stage of disease undergoing similar treatment regimes. The current staging system codifies the tumor extent based solely on anatomical information, which is insufficient for estimating prognosis for individual patients because it does not reflect the biological heterogeneity of cancer. Rapidly increasing findings on cancer biology provide prognostic information that complements and, in some cases, is more relevant than anatomical extent. Therefore, the discovery and application of molecular biomarkers that can be incorporated into the cancer staging system could improve the accuracy of prognostic prediction.
Ideal biomarkers should be easily accessible and noninvasive; therefore, there is great interest in circulating nucleic acids.[6-8] Circulating microRNAs (miRNAs) are stably detectable in serum or plasma, and their levels are reproducible and consistent among individuals of the same species.[9-11] miRNAs are small, noncoding RNAs that regulate a variety of biological processes. miRNAs originating from various tissues or organs can be released into the circulation during specific disease processes, which indicates the potential of circulating miRNAs as novel noninvasive diagnostic and prognostic biomarkers for a variety of diseases, including cancer.[12-20] Recent research has demonstrated that miRNAs are aberrantly expressed in NPC[21-23] and serum miRNAs have potential as biomarkers for the diagnosis of NPC. However, no study has evaluated the prognostic value of serum miRNAs in NPC.
In our study, we examined the level of serum miRNA in a large cohort of NPC patients to identify a serum miRNA signature associated with patient survival for the further development of personalized therapy for NPC patients.
Material and Methods
Our study was approved by the Ethics Committee of Sun Yat-sen University Cancer Center; all patients provided written informed consent for their specimens to be used. In total, 512 serum specimens from newly diagnosed, biopsy-proven, nonmetastatic NPC patients were collected at Sun Yat-sen University Cancer Center between January 2001 and December 2006. No patients had received any treatment, such as radiotherapy or chemotherapy, before blood sampling. The clinicopathological features of the patients are presented in Table 1.
|Training set (n = 256)||Validation set (n =256)||Combined set (n = 512)|
|High-risk group (%)||Low-risk group (%)||High-risk group (%)||Low-risk group (%)||High-risk group (%)||Low-risk group (%)|
|Characteristic||n = 128||n = 128||p-Valueb||n = 116||n = 140||p-Valueb||n = 244||n = 268||p-Valueb|
|Age (mean ± SD)||44.52 ± 11.25||46.68 ± 11.98||0.14a||47.06 ± 11.70||44.60 ± 11.32||0.09a||45.73 ± 11.51||45.59 ± 11.66||0.90a|
|Male||101 (79)||95 (76)||0.38||80 (69)||104 (74)||0.31||181 (74)||109 (74)||0.94|
|Female||27 (21)||33 (24)||36 (31)||36 (26)||63 (26)||69 (26)|
|WHO pathologic type|
|Undifferentiated nonkeratinizing||116 (91)||124 (97)||0.10||107 (92)||132 (94)||0.48||222 (91)||256 (96)||0.12|
|Differentiated nonkeratinizing||11 (9)||4 (3)||6 (5)||7 (5)||17 (7)||11 (4)|
|Keratinizing squamous cell||1 (1)||0 (0)||3 (3)||1 (1)||4 (2)||1 (0)|
|<1:80||14 (11)||18 (14)||0.75||15 (13)||22 (16)||0.67||29 (12)||40 (15)||0.55|
|1:80–1:320||73 (57)||70 (55)||67 (58)||83 (59)||140 (57)||153 (57)|
|≥1:640||41 (32)||40 (31)||34 (29)||35 (25)||75 (31)||75 (28)|
|<1:10||26 (20)||36 (28)||0.33||24 (21)||37 (26)||0.16||50 (20)||73 (27)||0.08|
|1:10–1:20||41 (32)||39 (30)||36 (31)||52 (37)||77 (32)||91 (34)|
|≥1:40||61 (48)||53 (41)||56 (48)||51 (36)||117 (48)||104 (39)|
|I||4 (3)||6 (5)||0.75||3 (3)||6 (4)||0.60||7 (3)||12 (4)||0.62|
|II||31 (24)||28 (22)||22 (19)||34 (24)||53 (22)||62 (23)|
|III||47 (37)||42 (33)||44 (38)||46 (33)||91 (37)||88 (33)|
|IV||46 (36)||52 (41)||47 (41)||54 (39)||93 (38)||106 (40)|
Two radiologists (L.Z.L. and L.L.) independently re-evaluated all MRI and CT scans, and any disagreements were resolved by consensus. TNM clinical stage was reclassified according to the 7th edition of the AJCC Cancer Staging Manual. All patients were treated with conventional two-dimensional radiotherapy, and 300 of the 378 (79.4%) patients with advanced disease (T3-T4 or N2-N3) also received platinum-based induction or concomitant chemotherapy. The median follow-up time for the whole group was 78.4 months (range: 6.7–133.1 months).
Sixteen patients were selected initially, and their serum miRNA profiles were determined using a microarray. Of these 16 patients, eight patients who had survived without relapse or metastasis for more than 5 years at the last follow-up were classified as the longer survival group; the other eight patients who survived less than 2 years because of relapse or metastasis were classified as the shorter survival group. Differentially altered miRNAs were validated by quantitative RT-PCR in the 16 samples from these patients.
All 512 samples were randomly divided into a training set (n = 256) and a validation set (n = 256) using computer-generated random numbers. The prognostic values of the four miRNAs were evaluated using RT-qPCR, and a miRNA signature was constructed using the risk-score method in the training set.
The miRNA signature was further validated in the validation set and the entire patient cohort.
Serum preparation and RNA extraction
Venous blood was incubated at room temperature for 1 hr, centrifuged at 500g for 10 min and then 10,000g for 30 min at 4°C to separate the serum. The supernatant was transferred to fresh RNase-free tubes and stored at −80°C until use. Total RNA was isolated from 400 µL of serum using TRIzol® LS reagent (Life Technologies, Carlsbad, CA) and precipitated with isopropanol containing GlycoBlue™ according to the manufacturer's instructions (Ambion, Carlsbad, CA). The RNA quantity and quality were assessed using a NanoDrop ND 1000 (Thermo Fisher, Boston, MA) and Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA).
Total RNA (100 ng) was labeled and hybridized to the Agilent human miRNA V12.0 microarray (Agilent Technologies). After hybridization, the slides were scanned using an Agilent Microarray Scanner, and the images were converted into intensity values using Feature Extraction software 10.7 (Agilent Technologies). After background subtraction, the data were imported into Gene Spring software 11.0 (Agilent Technologies) for quantile normalization and further analysis. A detectable miRNA was defined as a miRNA with positive microarray signals in >50% of the serum samples from either the longer or shorter survival groups. The microarray data have been deposited in the NCBI Gene Expression Omnibus under accession number GSE43160 (http://www.ncbi.nlm. nih.gov/geo).
Total RNA was first incubated for 10 min at 70°C in a 19 µL reaction volume containing 200 ng total RNA, 4 µL Bulge-Loop™ miRNA-specific RT primers (62.5 nM for miR-22, miR-572, miR-638, miR-1234 and miR-16, RiboBio, Guangzhou, China) and ddH2O and then reverse-transcribed in a 50 µL reaction volume by adding 1 µL M-MLV Reverse Transcriptase (200 U/µL, Promega, Madison, WI), 1 µL RNasin Plus RNase inhibitor (40 U/µL, Promega), 1 µL dNTP mix (10 mM, Promega), 10 µL 5× M-MLV buffer and 18 µL ddH2O with the following parameter values: 60 min at 42°C, 10 min at 70°C and stored at −20°C until use. Real-time PCR reactions were performed on the PRISM 7900HT system (Applied Biosystems, Carlsbad, CA) in a 20 µL reaction volume containing 2 µL reverse transcription product, 9 µL Platinum SYBR Green qPCR SuperMix-UDG reagents (Invitrogen, Carlsbad, CA), 2 µL PCR Forward Primer (5 µM, RiboBio), 2 µL Universal Adaptor PCR Primer (5 µM, RiboBio) and 5 µL ddH2O. The reactions were incubated at 95°C for 10 min, following by 40 cycles (95°C for 15 sec and 60°C for 30 sec) and then ramped from 60 to 95°C to obtain the melting curve. MiR-16 was chosen as the endogenous control; the coefficient of variation (CV) for the cycle threshold (Ct) values for miR-16 among all samples was small (CV = 0.038). All assays were performed in triplicate, and reactions without reverse transcription or RNA templates were used as a negative control. The levels of the miRNAs were calculated using the 2–ΔCT method.[18, 26]
Our study had a power of 80% to detect a survival failure hazard ratio (HR) of 2.29 (two-sided log-rank test, p = 0.05), assuming a 5-year overall survival (OS) rate of 60% in the high-risk group and 80% in the low-risk group. It was anticipated that 77 events were required from a total of 256 patients in the training set (128 per risk group).
The primary and secondary end points were OS and distant metastasis-free survival (DMFS), respectively, which were calculated from treatment to the date of death from any cause or first distant metastasis. Unpaired t-tests were used to identify miRNAs that were differentially altered between the shorter and longer survival groups (threshold difference of two-fold, p < 0.05). The risk-score method was used to construct a signature using differentially altered miRNAs.[18, 28] The relationships between the miRNA signature and various clinical characteristics were analyzed using Student's t-test, the chi-squared test or Fisher's exact test. The Kaplan–Meier method and log-rank test were used to estimate OS and DMFS.
Multivariate Cox regression analysis with a backward stepwise method was conducted to test for independent prognostic factors; the miRNA signature, age, sex, pathological type, Epstein–Barr virus (EBV) seromarkers and stage were used as covariates. Stratified analysis was performed to test whether the miRNA signature was associated with survival independent of stage. A prognostic score model was constructed by combining the four-miRNA signature and stage[23, 29]; receiver operating characteristic (ROC) curves were used to compare its prognostic validity with that of the stage-alone and the four-miRNA signature-alone models.
Potential target genes for miRNAs were predicted by three databases (TargetScan, PicTar and miRDB). A list of target genes was submitted to DAVID Bioinformatics Resources 6.7 (http://david.abcc.ncifcrf.gov) to acquire the gene annotation, and a graph of signaling pathways for NPC carcinogenesis in which the target genes may be involved was generated using the KEGG pathways program (http://www.genome.ad.jp/kegg). All statistical analyses were performed using Stata 10.0 software (StataCorp LP, College Station, TX) with two-tailed tests; statistical significance was defined as p < 0.05.
Identification of a four-miRNA signature associated with survival in the training set
The eight shorter survival NPC patients and eight longer survival NPC patients were well matched in terms of age, sex, pathological type, EBV seromarkers and clinical stage. A microarray was initially used to screen for differentially altered miRNAs between the shorter and longer survival groups; 16 serum miRNAs were differentially altered (fold change ≥ 2; p < 0.05; see Supporting Information Table 1). To validate the microarray data, we performed RT-qPCR for each of these 16 identified miRNAs, and found that only four miRNAs (miR-22, miR-572, miR-638 and miR-1234) were differentially altered between the shorter and longer survival groups (all p < 0.05).
All 512 NPC serum samples were randomly divided into a training set and a validation set. The four identified miRNAs were first quantified in the 256 samples from the training set, and univariate Cox regression analysis was used to analyze their associations with OS. As shown in Supporting Information Table 2, the levels of three miRNAs (miR-22, miR-572 and miR-638) were inversely associated with OS, and the level of miR-1234 was positively associated with OS. A risk-score method was used to construct a signature using these four miRNAs. Each patient was assigned a risk score based on a linear combination of the levels of the four miRNAs, weighted by their regression coefficients[18, 28] as follows: risk score = (0.146 × level of miR-22) + (0.288 × level of miR-572) + (0.182 × level of miR-638) − (0.272 × level of miR-1234).
Patients in the training set were divided into high- or low-risk groups using the median risk score (−2.90) as the cutoff point. There were no significant differences in clinical characteristics between the high- and low-risk groups (all p > 0.05, Table 1); however, patients with high-risk scores had poorer OS (HR, 2.54; 95% confidence interval (CI), 1.57–4.12; p < 0.001) and poorer DMFS (HR, 3.28; 95% CI, 1.82–5.94; p < 0.001) than those with low-risk scores (Figs. 1a and 1b).
Validation of the four-miRNA signature in the validation set and combined set
To validate the prognostic value of the four-miRNA signature, we quantified the levels of the four miRNAs in the 256 samples from the validation set, calculated the risk score for each patient using the same formula and classified the patients in the validation set into high- and low-risk groups using the same cutoff point that we used in the training set. Similarly, patients with high-risk scores were associated with significantly shorter OS (HR, 2.37; 95% CI, 1.47–3.82; p < 0.001) and shorter DMFS (HR, 3.31; 95% CI, 1.86–5.91; p < 0.001) than those with low-risk scores (Figs. 1c and 1d). The clinical characteristics of the high- and low-risk groups did not vary significantly (all p > 0.05; Table 1).
Furthermore, we combined all 512 patients into a combined set and found that patients with high-risk scores had shorter OS (HR, 2.40; 95% CI, 1.71–3.37; p < 0·001) and shorter DMFS (HR, 3.31; 95% CI, 2.18–5.02; p < 0·001) than those with low-risk scores (Figs. 1e and 1f). The clinical characteristics of the high- and low-risk groups did not differ significantly (all p > 0.05, Table 1). We also determined the distribution of risk scores, survival status and miRNA levels for the patients in the combined set. Patients with high-risk scores tended to have higher levels of the risky miRNAs and had more death than those with low-risk scores (Fig. 2). Similar results were observed in the training and validation sets (see Supporting Information Figs. 1 and 2). We further analyzed the prognostic value of each individual miRNA (see Supporting Information Results, Figs. 3–6 and Table 3).
The four-miRNA signature is associated with survival independent of stage
To investigate whether the four-miRNA signature could predict survival within stages, we performed stratified analyses in Stage II, III and IV patients from the whole patient cohort. Stage I patients were excluded, as no death or metastasis occurred in this subgroup. Stage II, III and IV patients with high-risk scores had poorer OS (p = 0.038, p = 0.006 and p < 0.001, respectively) and poorer DMFS (p = 0.045, p = 0.005 and p < 0.001, respectively) (Fig. 3). In addition, multivariate Cox regression analyses revealed that the four-miRNA signature and stage were independent prognostic predictors for OS (HR, 2.40; 95% CI, 1.71–3.37; p < 0.001 and HR, 2.78; 95% CI, 1.72–4.51; p < 0.001, respectively) and DMFS (HR, 3.31; 95% CI, 2.18–5.02; p < 0.001 and HR, 3.30; 95% CI, 1.81–6.01; p < 0.001, respectively) in the combined set. Similar results were also found in the training and validation sets (Table 2).
|MicroRNA signature (high- vs. low-risk)||2.54||1.57–4.12||<0.001|
|TNM stage (III–IV vs. I–II)||3.18||1.63–6.21||0.001|
|MicroRNA signature (high- vs. low-risk)||2.37||1.47–3.82||<0.001|
|TNM stage (III–IV vs. I–II)||2.66||1.32–5.35||0.006|
|MicroRNA signature (high- vs. low-risk)||2.40||1.71–3.37||<0.001|
|TNM stage (III–IV vs. I–II)||2.78||1.72–4.51||<0.001|
|Distant metastasis-free survival|
|MicroRNA signature (high- vs. low-risk)||3.28||1.82–5.94||<0.001|
|TNM stage (III–IV vs. I–II)||3.78||1.62–8.83||0.002|
|MicroRNA signature (high- vs. low-risk)||3.31||1.86–5.91||<0.001|
|TNM stage (III–IV vs. I–II)||2.97||1.27–6.94||0.012|
|MicroRNA signature (high- vs. low-risk)||3.31||2.18–5.02||<0.001|
|TNM stage (III–IV vs. I–II)||3.30||1.81–6.01||<0.001|
The four-miRNA signature adds prognostic value to the TNM staging system
We constructed a prognostic score model including only two independent prognostic factors, i.e., the miRNA signature and TNM stage, based on the data from the training set.[23, 29] The regression coefficient of the miRNA signature was divided by the regression coefficient of the TNM stage and rounded into an integer value to generate risk scores, which were used to calculate the cumulative risk score for each patient (see Supporting Information Table 4). ROC analysis showed that the model combining the miRNA signature and TNM stage had improved prognostic value for OS [area under ROC (AUROC): 0.69 vs. 0.60, p = 0.001; AUROC: 0.69 vs. 0.63, p = 0.008] and DMFS (AUROC: 0.71 vs. 0.60, p < 0.001; AUROC: 0.71 vs. 0.65, p = 0.005) relative to the TNM stage-alone model or miRNA signature-alone model in the training set (Figs. 4a and 4d). Similar results were also found in the validation (Figs. 4b and 4e) and combined sets (Figs. 4c and 4f).
On the basis of genome-wide serum miRNA profiling, we identified a four-miRNA signature significantly associated with survival in NPC, even after stratification by clinical stage. In addition, combination of the four-miRNA signature and TNM stage had a better prognostic value than TNM stage alone. These results suggest that serum miRNAs may play important roles in NPC development and progression, and may serve as potential biomarkers.
Precise prognostic predictions are crucial for making appropriate treatment choices for NPC patients. At present, the TNM staging system has been used for many years and is widely accepted as a prognostic predictor for NPC. In the clinic, NPC patients with the same stage who receive similar treatment regimes often show large variations in disease outcomes, which indicates that the current staging system may have reached its limitations for prognostic assessment. Molecular biomarkers may provide additional value for prognostic prediction.[4, 5] Circulating miRNAs are stably detectable in serum or plasma[9-11] and have potential as novel prognostic biomarkers in a variety of cancers[18-20]; however, little is known about the relationship between serum miRNAs and clinical outcomes in NPC. Therefore, we determined the serum miRNA profiles in a large cohort of NPC patients, and identified a four-miRNA signature that was significantly associated with patient survival. Compared to patients with low-risk scores, NPC patients with high-risk scores had reduced OS and increased distant metastasis, verifying the prognostic value of serum miRNAs. This miRNA signature may enable the identification of patients who may benefit from more aggressive treatment and therefore improve survival in NPC patients.
In addition, we found that the four-miRNA signature could predict survival independent of clinical stage. The TNM staging system classifies disease extent based mainly on anatomic location; however, the serum miRNA signature can reflect the biological features of NPC. Thus, the serum miRNA signature, in conjunction with the TNM staging system, has the potential to more effectively evaluate patient prognosis and further select adequate treatment strategies. Therefore, we constructed a prognostic score model that included two independent prognostic factors,[23, 29] the four-miRNA signature and TNM stage, and demonstrated that the combination of the miRNA signature and TNM stage had a better prognostic value than TNM staging alone, indicating that the miRNA signature could add prognostic value to the TNM staging system. As a result, NPC patients with the same clinical stage could be distinguished into different risk groups with favorable or unfavorable survival and could be systemically treated with different approaches or intensities to improve their disease outcomes.
The standard treatment for NPC patients with early-stage disease is radiotherapy, and the treatment for patients with advanced disease is chemoradiotherapy. The local control rate has exceeded 90%; however, ∼30% of patients die because of distant metastasis, which has become the predominant pattern of treatment failure and cause of death.[30-32] This pattern of failure indicates that a subset of patients do not benefit from the current therapeutic strategies and may receive toxic therapy unnecessarily. The miRNA signature could classify NPC patients into high-risk and low-risk groups so that more intensive, targeted therapeutic strategies can be designed for patients with high-risk scores. New chemotherapeutic or targeted agents, such as docetaxel, cetuximab and sorafenib, have demonstrated clinical efficacy and acceptable safety in neoadjuvant chemotherapy and palliative settings for NPC patients.[33-35] Recent evidence also indicates that the reduction or replacement of specific miRNA may have therapeutic effects.[36, 37] The four miRNAs identified in our study potentially target 709 genes involved in 15 signaling pathways, such as focal adhesions, MAPK signaling and ErbB signaling, most of which are associated with invasion, metastasis and proliferation and are involved in NPC development (see Supporting Information Fig. 7). Therefore, further investigation focusing on new agents and the function of miRNAs may provide novel approaches and targets for the management of NPC.
Determination of the levels of a small number of miRNAs in serum using RT-qPCR, as shown in our study, may represent a clinically applicable, noninvasive procedure. However, there are some limitations in our study. First, it should be noted that the four-miRNA signature was identified from Chinese patients at a single medical center. Our findings therefore warrant further studies at other centers and, ideally, in different ethnic populations. Second, although studies suggest that serum miRNAs may be selectively released from tumor cells[12, 13] via mechanisms involving tumor-derived exosomes or microvesicles,[38, 39] the origins and mechanisms leading to the generation of serum miRNAs remain unclear. Third, several investigations have shown that serum miRNAs may potentially play biological roles as mediators of intercellular communication[38-40]; therefore, characterizing the functions and mechanisms by which miRNAs regulate survival in NPC may lead to the application of these serum miRNAs in the clinic and provide additional targets for NPC treatment.
In conclusion, we identified a serum miRNA signature that was significantly associated with patient survival in NPC. This serum miRNA signature can add prognostic value to the TNM staging system, which may lead to more personalized therapy. Multicenter, large-scale, prospective studies are necessary to validate our findings before this signature can be used in the clinic, and further investigation of the roles of these miRNAs is also necessary.
- 2Cancer incidence in five continents, vol. IX. Lyon, France: International Agency for Research on Cancer, 2007., , , et al.
- 3AJCC cancer staging manual, 7th edn. New York: Springer, 2010., , , et al.
- 27Sample size calculations in clinical research. New York: Marcel Dekker, 2003., , .
Additional Supporting Information may be found in the online version of this article.
Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.